1: \begin{abstract}

2: We discuss a modification to Random Matrix Theory eigenstate statistics,

3: that systematically takes into account the non-universal short-time behavior of chaotic systems.

4: The method avoids diagonalization of the Hamiltonian, instead requiring only a knowledge of short-time

5: dynamics for a chaotic system or ensemble of similar systems.

6: Standard Random Matrix Theory and semiclassical predictions are recovered in

7: the limits of zero Ehrenfest time and infinite Heisenberg time, respectively.

8: As examples, we discuss wave function autocorrelations and cross-correlations,

9: and show that significant

10: improvement in accuracy is obtained for simple chaotic systems where comparison can be made with brute-force diagonalization.

11: The accuracy of the method persists even when the short-time dynamics of the system or ensemble is known only in a classical approximation. Further improvement in the rate of convergence is obtained when the method is combined with the correlation function bootstrapping approach introduced previously.

12: \end{abstract}

13: